Shiguang Wen

An Improved Algorithm of Hand Gesture Recognition under Intricate Background

This paper presents an integrated algorithm of YCbCr-Nrg, Double Color-Spatial Model and Background Model to resolve the problem that single skin-color model is obstructed by near kin color. This segmentation method is realized by the fusion of muti-feature. Based on the good describing ability of Fourier Descriptors algorithm and the good self-learning ability of BP neural network, an improved algorithm of hand recognition is presented and carried out. Results show that this algorithm is robustness for hand gesture recognition under intricate background.

A novel interactive method of virtual reality system based on hand gesture recognition

Hand gesture is a natural and intuitive interactive method. This paper presents a novel interactive method of virtual reality system based on hand gesture recognition. The hand gesture segmentation method is proposed based on building complexion model by Gaussian distribution and the background model by automatically update the background parameters to improve the ability of adaptation environment. According to the good describing ability of Fourier Descriptor and the good self-learning ability of BP neural network, an improved algorithm of hand recognition is presented. Experiment result indicates that this method is flexible, realistic and exact, and fit for many virtual reality systems.

EEG based automatic left-right hand movement classification

As a research focus in bio-rehabilitation field, Brain-Computer Interface that can efficiently improve or recover self-care ability of disabled people has obtained fast development in recent years. In this paper, the phenomenon of event-related de-synchronization in EEG was pre-processed and features were extracted by using common space pattern filtering algorithm. Higher classification accuracy was then obtained by employing C-SVM for left-right hand movement discrimination. Since the effectiveness and simplicity of the proposed scheme, it can be used as an on-line classification algorithm in BCI system.

Realtime gait kinematics classification using LDA and SVM

Gait analysis is an important part of intelligent prosthesis researching. The automatic segmentation and classification of gait kinematic signal could help intelligent prosthesis to get better control performance. Much feature extraction method was employed by researchers, but it could still be improved. The wavelet based filter is adopted in this paper to segment the gait kinematics data, and LDA/SVM is employed to improve the accuracy of recognition. Result shows better performance than that using traditional method.

Nao humanoid robot gait planning based on the linear inverted pendulum

The gait planning of humanoid robot Nao is solved based on the linear inverted pendulum (LIPM) in this paper. The step length and period is selected according to the structure of Nao. LIPM model and ZMP is applied in the planning. An ideal gait data is obtained through simulation experiment, demonstrating the effectiveness of the LIPM model in the gait planning of the biped robot.

Gait recognition based on the Fast Fourier Transform and SVM

Gait recognition is a new type of biometric identification technology. In the preliminary study of biped robot with heterogeneous legs, we detect road information by classify of subjects gait data by measurement using gyroscope. Then we can make bionic legs movement to follow artificial legs movement perfectly. So gait recognition is a very important step in it. This paper adopted the Fast Fourier Transform to the feature selection of time-series sequence of measurement gait data, and reduces the characteristic number of the samples into two. Then we detected the situation of road by gait recognition of different gait data on the different roads based support vector machine. The experiment shows a good result, and the method reduces the computing time significantly.

Walking gait generation using linear inverted pendulum model for biped robot with heterogeneous legs

As a general advanced plateform of performance test and quantitative estimation for prosthetic leg, biped robot with heterogeneous legs (BRHL) simulates the amputee-prosthesis coupling system. To generate the walking gait trajectories of artificial leg, linear inverted pendulum model combining with ZMP stability criterion was employed and the reference angle of joints was calculated by establishing and solving robot inverse kinematics. To eliminate the effect of servo mechanism hysteresis, preview control was used to optimize the CoG trajectories. Experiment results indicate the effectiveness of the proposed scheme.

Intelligent bionic leg motion estimation based on interjoint coordination using PCA and RBF neural networks

It has been a challenging endeavor for amputee to coordinate harmoniously with his/her artificial limb. In this paper, a novel scheme of real-time motion estimation for Intelligent bionic leg based on interjoint coordination is proposed. To measure the gait during walking, inertial sensors are mounted on the CoGs of bilateral thighs and shanks to acquire angular velocities of lower limbs of subjects. For the existence of linear correlation between bilateral kinematics in healthy symmetrical human gait, principle components analysis is employed to model the interjoint coordination and is used to estimate the knee joint angle of intelligent bionic leg from body motion of amputee. To improve the presicion of motion estimation further, RBF neural networks are used to optimally calculate the knee joint angle. Simulation and experimental results demonstrate the effectiveness and correctness of the proposed scheme.

Study of gait symmetry quantification and its application to intelligent prosthetic leg development

Human bipedal walking is symmetrical, stable and efficiency optimal. To clarify the essential relationship in human gait, accelerometers and rate gyroscopes are mounted on waist and lower extremities of subject to acquire kinematics information during walking. By employing Principal Component Analysis (PCA), kinematics signals of one side are reconstructed using that of opposite side. The reconstructed signals match the acquired ones very well, which verifies the existence of strong linear correlations in healthy human gait. To evaluate the symmetry of gait quantitatively, several classical temporal-spatial features for gait calculated by algorithms involving autocorrelation function are selected for construction of quantitative criterions by combining with symmetry indices. For the validation of gait symmetry quantification scheme, level walking experiments including healthy human subjects and human-intelligent prosthetic leg coupling system are conducted. Results indicate the correctness and effectiveness of the proposed scheme and gait symmetry analysis shows great potential in the field of bio-medical rehabilitation.

An on-line gait generator for bipedal walking robot based on neural networks

An on-line gait synthesis scheme for a bipedal walking robot is proposed. To realize efficient and human-like gait, MTi sensors were mounted on the lower limb of human subject to acquire kinematics information that can be integrated to the angular changes of hip and knee joints during walking. The time series angles were normalized and then sampled by cubic spline interpolation. By employing discrete-time Fourier Series, the samples were extracted into features, and further dimensionally reduced by using PCA to simplified features. By using ANNs, the nonlinear functional relations between gait parameters (i.e. cadence and stride) and simplified features that can be used to reconstruct angles of hip and knee joints were established. Walking experiments of a biped robot at slow, intermediate and fast speeds were conducted to validate the effectiveness of the proposed scheme. The results indicate that the synthesized gait is smooth, efficient and human-like. The proposed scheme can on-line generate the reference gait that covers a wide speed range for bipedal walking of robot.